Diagnostics for linear models with Gaussian process random effects

Stationary Gaussian processes are often included as latent processes in linear models to incorporate the effects of time and space. The recent paper by Bose, Hodges and Bannerjee (2018, Biometrics) proposed some interesting methods of doing diagnostics for such models. Specifically, they proposed methods for understanding how data determine estimates of the parameters in these models and for exploring the contribution of covariates to explaining the response. The project is to understand and explore these methods.

This project will be co-supervised with Dr Francis Hui